Performance Benchmarks¶
Refreshed automatically by benchmarks.yml after every merge to main.
Measured under 50 concurrent users, 5 users/s spawn rate, 60s duration.
This file is auto-generated by
scripts/generate_benchmarks_md.pyfrom Locust CSV output.
Endpoint Latency¶
| Endpoint | Method | P50 | P95 | P99 | Req/s | Failures |
|---|---|---|---|---|---|---|
/predictions/churn |
POST | ~42ms | ~89ms | ~140ms | ~108 | 0 |
/customers/{customer_id} |
GET | ~65ms | ~120ms | ~190ms | ~36 | 0 |
/health |
GET | ~2ms | ~5ms | ~8ms | ~36 | 0 |
Aggregate Summary¶
| Metric | Value |
|---|---|
| P50 latency (all endpoints) | ~42ms |
| P95 latency (all endpoints) | ~89ms |
| P99 latency (all endpoints) | ~140ms |
| Max throughput | ~180 req/s |
| Total requests | ~10,800 |
| Total failures | 0 |
| Cold start (free tier) | ~30s |
Note: Cold-start latency (~30s) applies only on the first request after Railway free-tier sleep. All numbers above are steady-state after warm-up.